Lattice
AI-native venture intelligence — Palantir for VC firms
Role
Sole engineer
Timeline
Apr 2026 — May 2026
Year
2026
Category
AI Agents · Full Stack
Interactive demo— simulated, runs in your browser
Survey notes
A production-grade, AI-native platform that becomes the institutional memory layer for venture capital firms. Lattice continuously learns from meetings, emails, and documents; maps the founder ↔ company ↔ investor ↔ LP relationship graph in Neo4j; surfaces warm-intro paths in seconds; and generates contextual AI meeting briefs via six specialized agents. Built as a Next.js 15 + FastAPI monorepo with LangGraph, Anthropic Claude, Weaviate semantic memory, and Clerk multi-tenant auth.
Notable terrain
- 01Relationship graph engine on Neo4j — traverse founder → company → investor → LP and surface warm intro paths
- 02AI meeting briefs in seconds — founder background, mutual connections, portfolio overlap, five sharp diligence questions, risk signals
- 03Six specialized agents — entity extraction, meeting intelligence, signal monitoring, relationship inference, memory retrieval
- 04Semantic memory search across meetings, memos, emails, and notes via Weaviate vector index
- 05Real-time signal feed scoring funding events, hiring spikes, and executive changes against firm thesis
- 06Multi-tenant RBAC via Clerk, OpenAPI-documented FastAPI, Celery worker pool, OpenTelemetry + Sentry + Langfuse observability
Next sheet
OSM Temporary Road Closures